Field Notes / 01  ·  Policy

Writing an AI acceptable-use policy your board will actually pass.

Most district AI policies die in one of two ways: they ban everything and get ignored, or they say nothing and get challenged. Here's the structure that survives a board vote — and the first year after it.

Matthew Kelso, Ed.D  ·  July 2026  ·  6 min read

A superintendent showed me a draft AI policy last spring that was nine pages long and mentioned students twice. It had been assembled from three vendor templates and a neighboring district's handbook. The board tabled it. They were right to.

A policy the board will pass — and defend at a public meeting — has to answer the questions parents actually ask. Who can use these tools? What happens to my kid's data? Who decided this? The structure below comes from policies that made it through real board votes, and from watching Ohio put a hard deadline on the question: every public district there must publish a comprehensive AI policy by July 1, 2026. Other states are following. The era of "we'll get to it" is over.

Start with who decides, not what's banned

The strongest policies establish an AI review committee before they say a word about tools. Technology, teaching staff, student services, and someone with data-privacy responsibility — that's the minimum table. Every AI system that touches staff or student data goes through that committee before it enters a classroom. This one structure does more than any list of banned tools, because the list will be stale in a semester and the committee won't.

The six sections that need to exist

Appropriate use, by role. What teachers may do (lesson planning, differentiation, feedback drafts), what students may do at each grade band, and what nobody may do. Several state frameworks now draw the same lines: no open chatbots in PreK-2, curriculum-embedded AI only in grades 3-5, structured and supervised use in middle school. Borrow those lines — they exist because someone already fought about them.

Human decision authority. One sentence your board attorney will love: no final decision about discipline, placement, or evaluation is made solely by an automated system. A person decides; a tool assists.

Data privacy. Name the laws — FERPA, COPPA, your state's student data privacy statute — and name the mechanism: no AI tool is used with student data until a data-processing agreement is signed. If you want the detail on what that agreement must contain, that's its own field note.

Academic integrity, stated positively. Define what assisted work is allowed and how students disclose it. Districts that only wrote "AI plagiarism is prohibited" spent the next year adjudicating what plagiarism means when the tool wrote a third of the sentence. Disclosure norms age better than prohibition lists.

Parent notice and consent records. An annual notice of what AI systems the district uses, folded into the student technology agreement, with records kept of who consented to what. This is the section that turns a hostile public-comment session into a quiet one.

Review cadence. The policy names its own expiration: annual review by the committee, with educators, parents, and students consulted. A policy that can't change is a policy that gets abandoned.

What kills policies at the board table

Length, borrowed language, and missing money. Nine pages of vendor prose signals that nobody in the district owns the thing. Aim for three to four pages a principal can actually apply. And if the policy requires training — it should — the board will ask what that costs. Have the number ready, or the vote becomes a budget discussion you didn't prepare for.

The other killer is skipping the teachers. A policy drafted in the central office and announced at a staff meeting produces quiet non-compliance at scale. The committee structure fixes this too: teachers who helped draw the lines defend the lines.

A realistic timeline

Committee formed in month one. Draft in month two, built on your state's guidance document if one exists — most states have published one. Legal review and a public comment window in month three. Board vote in month four, training before the semester starts. Districts that compress this into three weeks before a board meeting get the nine-page tabled draft. Districts that stretch it past a year get passed by their own students.

If you want a second set of eyes on a draft — or a facilitator for the committee's first meeting — that's work we do. Bring the draft; we've read enough of them to know where yours will get stuck.

Want the starting framework?

Get the AI Readiness Checklist